installer-packs

Installer Packs

comfyui-mcp ships installer packs under packs/. Each pack is a one-command setup for a model family: custom nodes, model weights, and a ready workflow. A single manifest.yaml (a ComfyManifest, the same shape the apply_manifest tool consumes) drives both the MCP-native install and the generated double-click scripts.

packs/<name>/
  manifest.yaml         # custom_nodes + models (url → local_path) — source of truth
  pack.yaml             # metadata: workflow, family, VRAM, sources, notes
  workflow.json         # the graph to load
  install-windows.bat   # GENERATED — never hand-edit
  install-runpod.sh     # GENERATED — never hand-edit

Installing a pack

• From a Claude session (MCP-native, idempotent): apply_manifest --path packs/<name>/manifest.yaml (requires COMFYUI_PATH). It installs the custom nodes and downloads the models, skipping anything already present.
• One-click for non-MCP users: run packs/<name>/install-windows.bat (or install-runpod.sh) from a ComfyUI root. Then load the pack's workflow.json.
• After install, check the pack's pack.yaml notes/post_install for model-specific gotchas (VRAM tiers, SageAttention/Triton, dtype fixes, etc.).

Building or deriving a new pack

Two sources of ground truth, in order of preference:

1. An upstream installer (*-MODELS-NODES_INSTALL.bat / .sh). Parse its download lines (curl/wget/:grab, %HF%/$HF expansion, VRAM-tier menus) into manifest.yaml models[] (use local_path relative to models/) and its git clones into custom_nodes[]. Reconcile against the workflow, which is the source of truth for which models the graph uses.
2. A workflow alone (no installer). Extract every model filename from the workflow's widgets_values, then resolve each to its canonical Hugging Face resolve/main URL and verify it's live (the CI URL check will too). Never invent a URL. Flag what you can't resolve. Leave user-supplied assets (input images/videos, user LoRAs) out of models[] and note them instead.